鷹群掠食演算法於全域最佳化設計之應用

Abstract

The biological behaviors in life have usually inspired the development of creative evolution computation algorithms, such as Ant Colony Optimization and Particle Swarm Optimization. According to eagle's four distinctive foraging behaviors, searching, exploring, striking，and killing, we proposed a population-based evolution computation technique called as Eagle-Foraging Algorithm (EFA) in the paper. The developed algorithm is completed under the frameworks of adopting the Particle Swarm Optimization (PSO), a space identification scheme, and a local search technique of the Hooke-Jeeves method. Eight benchmark test problems with different functional characteristics have been selected to validate the proposed EFA performance. In each tested problem, two respective 10 and 30 variables were used in the computation. The results demonstrate that the EF A can secure the solutions of all eight benchmark problems with high performance comparing several variants of PSO in references [13, 30]. Finally, in application, the EFA has been employed to optimize the structural design of a cantilevered beam.近年來，於演化計算領域中，仿生物行爲開發演算法如蟻群演算法與粒子群演算法越趨普遍。本文根據生態學者觀察食魚鷹掠食行爲，搜索、探親、俯衝入水與掠取魚獵物等四個步驟，發展一套族群式進化演算法，稱之爲鷹群掠食演算法(eagle-foraging algorithm, EFA)。本演算法架構整合粒子群演算法(particle swarm optimization, PSO)、空間鑑定法(space-identification scheme, SIS)W區域搜尋法(local search method)。 本文首先選取8種標竿函數問題以驗証本演算法之可靠性與效益性。測試結果顯示，於10與30維的問題之全域最佳解均可求得，除了有相當好的搜尋性能，而且比文獻[13, 30]比較佳。最後，本文將鷹群掠食演算法應用懸臂樑結構工程問題之最佳化設計。